Almost 20 years ago, Warren Sanborn predicted that ventilators today would “… report the patient’s metabolic state; manage oxygen delivery; calculate cardiac output, synchronize breath delivery with cardiac cycle to maximize cardiac output…and perform all these functions automatically or at least presenting consensus-based advisory messages to the practitioner….” 17
Some of these ideas were never developed commercially. Some were tried and abandoned. Some, have evolved beyond Warren’s broad vision.
There are three basic ways to improve ventilators in the future. First, just like computer games, ventilators need to improve the operator interface constantly. Yet very little research has been done to call attention to problems with current displays. 34 , 35 We have come a long way from using a
crank to adjust the stroke of a ventilator’s piston to set tidal volume. The operator interface must provide for three basic functions: allow input of control and alarm parameters, monitor the ventilator’s status, and monitor the ventilator– patient interaction status. We have a long way to go before the user interface provides an ideal experience with these functions.
Second, the weak link in the patient–ventilator system is the patient circuit. We buy a $35,000 ventilator with state-of-the-art computer control, and then we connect it to the patient (priceless) with a $1.98 piece of plastic tub- ing that is subject to filling with condensate from a heated humidifier whose design has not changed appreciably in 20 years. The resistance and compliance of the deliv- ery circuit make flow control and volume delivery more
difficult. It is like buying a Ferrari and putting wooden wheels on it. In the future, water vapor should be treated like any other desirable inhaled gas constituent (e.g., air, oxygen, helium, or nitric oxide) and metered from within the ventilator. The inspiratory part of the patient circuit should be a sterile, insulated, permanent part of the ven- tilator right up to the patient connection, which can be a disposable tip for cleaning purposes. The gas should be delivered under high pressure as a jet to provide not only conventional pressure, volume, and flow waveforms but also high-frequency ventilation. The jet also can be used to provide a counterflow PEEP effect, eliminating any need for an exhalation–valve system. The disposable tip could be designed to house disposable sensors and would be the only part of the circuit to be exposed to the patient’s exhaled gas. If ventilator manufacturers saw themselves as providers of the entire system, instead of letting third par- ties deal in plastic connecting tubing, I think we would see a huge evolutionary step in ventilator performance, better patient outcomes, and potential savings in labor costs for providers.
Third, the most exciting area for development probably is in the intelligence that will be built into future ventila- tor control circuits. The real challenge in closed-loop con- trol of ventilation is defining, measuring, and interpreting the appropriate feedback signals. If we stop to consider all the variables a human operator assesses, the problem looks insurmountable. Not only does a human consider a wide range of individual physiologic variables, but there are the more abstract evaluations of such things as meta- bolic, cardiovascular, and psychological states. Add to this the various environmental factors that may affect opera- tor judgment, and we get a truly complex control problem ( Fig. 2-13 ).
I would like to speculate now about a response to this challenge. The ideal control strategy would have to start out with basic tactical control of the individual breath. Next, we add longer-term strategic control that adapts to changing load characteristics. Mathematical models could provide the basic parameters of the mode, whereas expert rules would place limits to ensure lung protection.
Next, we sample various physiologic parameters and use fuzzy logic to establish the patient’s immediate condition. This information is passed on to a neural network, which would then select the best response to the patient’s condition.
The neural network ideally would have access to a huge database comprised of both human expert rules and actual patient responses to various ventilator strategies. This arrangement would allow the ventilator not only to learn from its interaction with the current patient but also to con- tribute to the database.
Finally, the database and this ventilator could be net- worked with other intelligent ventilators to multiply the learning capacity exponentially ( Fig. 2-14 ). Whatever the future brings, it seems clear that ventilators will have more intelligence built in to increase patient safety and decrease the time required to provide care.
Set-point Adjustment Environment Operator Alarms Flow Pressure Bronchospasm Underlying disease Strength/Endurance Neural control Auto-PEEP Metabolic state Acid–base state Cardiovascular state Psychological state Drugs Ventilator Time Cost Triage priority Experience Pressure (PIP and PEEP)
Volume Frequency FiO2 Disturbances Patient Pressure Volume Flow Resp rate Heart rate PeCO2 PaO2 FiO2 SpO2 P0.1
FIGURE 2-13 The challenge of total computer control of mechanical ventilation. Solid arrows depict signals that have been used at least experimen- tally. Dotted arrows represent potential feedback signals. (Reproduced, with permission, from Chatburn RL. Computer control of mechanical ventila- tion. Respir Care. 2004;49:507–515.)
Human experts Expert rules Ventilator Optimization models Registry Database Prior experience Fuzzy logic Competitive neural network Intelligent Strategic control Disturbances control
Networked ventilators Flow
Pressure Tactical control Patient Determine patient condition Determine best rules
FIGURE 2-14 A potential approach to the challenge of fully automated control of mechanical ventilation. (Reproduced, with permission, from Chatburn RL. Computer control of mechanical ventilation. Respir Care. 2004;49:507–515.)